Table 6 The average quality measures of each algorithm performed on the Q and H resolution training sets images
From: Real-time CUDA-based stereo matching using Cyclops2 algorithm
Algorithm | Bad 4.0, % | Invalid, % | Average error, pixel |
|---|---|---|---|
Quarter resolution dataset | |||
MD | 89.0 | 0.0 | 61 |
RMA | 77.4 | 0.2 | 41 |
MNS1 | 46.4 | 0.0 | 27 |
MNS2 | 72.9 | 0.0 | 38 |
MNS3 | 72.9 | 0.0 | 38 |
SGBM | 12.1 | 6.6 | 6 |
Half resolution dataset | |||
MD | 89.4 | 0.0 | 61 |
RMA | 78.3 | 0.1 | 42 |
MNS4 | 47.5 | 0.0 | 38 |
MNS5 | 76.5 | 0.0 | 43 |
SGBM | 8.7 | 9.7 | 7 |